Application of symbolic machine learning to the recognition of texture concepts

The authors present an approach to the texture recognition problem that deals with noisy learning and testing data. The method incorporates symbolic machine learning to acquire texture descriptions. Then, these descriptions are optimized in order to remove some noisy/imperfect components. The' authors present methodology and experimental results showing the increase in system recognition effectiveness when optimization of texture descriptions proceeds continuously. Such a matching of partial concept prototypes with test data gives recognition characteristics obtained for different concept optimization degrees. Then, the dynamics of these characteristics are used to make the recognition decision.<
>